MCP App Store
by KeenEthics
ServGround icon

ServGround MCP Server for ChatGPT

By ServGround

Overview

ServGround is the operations platform for field-service and environmental businesses: service requests, proposals, contracts, invoices, projects, reports and payments in one workspace. This plugin lets a workspace owner ask ChatGPT questions about their own business in plain language, such as "What's outstanding on invoices this month?" or "Which service requests came in this week?" - Read-only. It cannot create, change, send or delete anything in ServGround. - Owner-approved. Only a workspace owner can connect it, and they choose which areas ChatGPT may read. - Privacy option. With redaction on, client and contact names become initials, emails and phones are masked, and street addresses and parcel numbers are removed. Free text such as descriptions and notes is sent as written. - Revocable at any time from Settings > Integrations > Connected AI apps. A ServGround account with owner access is required.

How to connect

  1. ChatGPT: Open the ServGround listing in the ChatGPT app directory and choose Connect. ChatGPT listing ServGround exposes 28 tools here.

Tools

analyze_intake

ChatGPT
Where new work comes from for a date window: service requests broken down by how they arrived (source), how the client heard about the business (referral source), and current status.

analyze_job_hours

ChatGPT
How long jobs took across a period: count finished, captured hours, average/median/total time on site, longest and shortest jobs, estimate accuracy, optionally grouped by technician, client or month.

fetch

ChatGPT
Fetch one document by the id search returned — a document number such as SR-2026-273, INV-2026-014, PROP-2026-101, CON-2026-004, RPT-2026-008, PRJ-2026-031 or PMT-2026-002.

get_business_profile

ChatGPT
The business's own details: name, what it does, phone numbers, email, websites, the link to its request form, address, and the states and counties it serves. For the list of services the business offers use list_services.

get_business_summary

ChatGPT
CALL THIS FIRST for any overview, financial-health, or pipeline question. One scoped rollup: financials (collectedRevenue=cash in, billedRevenue, outstanding balance, net), invoices (paid/unpaid), proposals (total/accepted), contracts (total/signed), reports (total/published=sent), service requests (total/expedited), projects (total/assigned/unassigned), fieldTeam (active technicians + busiest), services (catalog total/active), team (office staff total), and needsReply (counts of ACTIVE SRs/invoices/reports with an outstanding client message awaiting a reply — terminal items excluded; list them with list_*?needsReply=true). Only includes the domains the credential is scoped for. Optional from/to date window. NOTE: this is a lightweight rollup — counts are by created_at (NOT event/paid date) and financials are top-line only. For an exact expense/fee BREAKDOWN, or a "collected/paid in this period" figure, call get_financials instead.

get_cash_flow

ChatGPT
Money in and money out by month or quarter for a date window, from the same figures as the Income and Expenses pages. Returns a chart specification with exact totals.

get_document

ChatGPT
One proposal, contract or report in full by its number: lines and totals, deposit and invoicing setup, status and dates, and what it is linked to. For an invoice use fetch with its INV number.

get_financials

ChatGPT
USE FOR every money TOTAL — income, revenue, expenses, fees, net/profit, outstanding/AR, refund totals. Exact SQL that MATCHES the Finances dashboard; NEVER sum a list tool (list_payments/list_income/list_expenses) to get a total. Exact revenue/expense rollup that MATCHES the merchant's Finances dashboard (Income + Expenses pages). INCOME: income = { total (Total Income = collectedRevenue + externalIncome), collectedRevenue (Platform Payments — SUCCEEDED card payments, excludes refunded), externalIncome (External Income — manual Zelle/cash/check entries, often 0), yourEarnings (= collectedRevenue − merchant payment fees, the "Your Earnings" card) }. Never report externalIncome alone as total income. EXPENSES: expenses = { total (= paymentFees + businessExpenses), paymentFees (all merchant-absorbed payment fees), servgroundFees (the ServGround/platform fee the merchant pays), processingFeesAbsorbed (Stripe processing fee the merchant absorbs — usually 0 because the client pays the surcharge), processingFeesCollectedFromClients (the surcharge CLIENTS paid — INFORMATIONAL, NOT a merchant expense), businessExpenses (manual expenses table, often 0) }. net = income.total − expenses.total. billedRevenue = invoices issued; outstanding = current unpaid balance (NOT date-filtered). refunds = money RETURNED to clients: { total, count = refund transactions } — for the per-INVOICE full/partial breakdown + names call list_invoices?refunded=full|partial. A refund (money out) is NOT a partially-PAID invoice (client still owes). For PER-PAYMENT fee detail (each payment's processing fee, ServGround fee, net, method) use list_payments; for external income rows use list_income. groupBy=month adds an exact monthly trend. Dates: collectedRevenue/fees on payment date, externalIncome on income date, refunds on refund date.

get_project

ChatGPT
One job in full by its project number (PRJ-YYYY-NNN): schedule, assigned crew, location, linked request/proposal/contract/invoice/reports, estimated vs actual hours, time on site and days taken. Its value is shown to owners only.

get_review_summary

ChatGPT
Client review requests for a date window: how many were sent, opened and rated, and how the ratings break down.

list_clients

ChatGPT
Client rows or an exact status rollup (groupBy=status). Each row includes status, portal_access_granted (true = the client portal is active for them) and sms_consent (true = they opted in to SMS). When the credential has redaction enabled, names (company names included) show as initials, email and phone are masked and addresses are removed; status, portal_access_granted and sms_consent are always shown.

list_contracts

ChatGPT
Contract rows (status, amount, sent/signed/declined dates, client_name) or an exact status/month rollup (groupBy=status|month). Includes client view tracking (first_viewed_at, view_count) — tells whether the client has seen the contract yet. Use to track the signing pipeline.

list_expenses

ChatGPT
USE ONLY to itemize manually-entered business expenses (often empty). This EXCLUDES payment/platform fees — so it is NOT total expenses. For total expenses (which include fees) use get_financials (expenses.total); for per-payment fee detail use list_payments. MANUAL business-expense records ONLY (the expenses table — often empty; the "Business Expenses" card). Does NOT include payment processing/platform fees. For TOTAL expenses (incl. fees) use get_financials (expenses.total); for the per-payment ServGround/processing FEE detail use list_payments. Use this only to list/break down manually-entered business expenses (groupBy=category|month).

list_field_team

ChatGPT
USE FOR field workers = technicians (roster + who is busy/free). For office staff (owner/member/coordinator) use list_team instead. Field workers = technicians (same thing). Roster + workload: every active technician (name, number) with their active job count/value in the window, sorted most-work-first (idle techs show 0 jobs). Answers "who are my field workers / technicians", "who is assigned the most", "who is free". Pass includeInactive=true to include deactivated technicians. No technician contact PII is exposed. (For OFFICE staff — owners/members/coordinators — use list_team instead.)

list_income

ChatGPT
USE ONLY to itemize EXTERNAL/manual income rows (Zelle, cash, check, Venmo, wire). This is NOT total income — for that use get_financials (income.total). Card payments live in list_payments, not here. EXTERNAL INCOME — money recorded OUTSIDE Stripe (Zelle, cash, check, Venmo, wire): the Income page "External Income" tab. Rows: incomeNumber, date, payerName, amount, method, category, description. groupBy=month for a trend. Distinct from Platform Payments (card payments → list_payments). For the income TOTAL use get_financials (income.externalIncome).

list_invoices

ChatGPT
Invoice rows (amounts, status, due/paid dates, client_name, needs_reply) or an exact status/month rollup (groupBy=status|month). Filter by status. Pass needsReply=true to return ONLY active invoices with an unanswered client message (excludes cancelled). To FIND REFUNDED INVOICES pass refunded=any|full|partial — it returns only invoices that had a refund (money returned) and attaches a refund { amount, type } to each row (full = whole collected amount returned; partial = some returned). Includes client view tracking (first_viewed_at, view_count) and needs_reply=true. NOTE: a partially-PAID invoice (status partially_paid, amount_paid < total_amount) is NOT a refund — different concept.

list_maintenance_jobs

ChatGPT
Recurring job series: repeating visits for a client, how often, when the next is due, who is assigned and whether the series is still running.

list_maintenance_plans

ChatGPT
Recurring payment plans: who is on one, what it charges (including the client fee), when it next charges, and whether it is billing, waiting to be set up, failing or stopped. Do not sum these into recurring revenue.

list_payments

ChatGPT
USE to itemize individual card/platform payments (per-payment fee detail); defaults to succeeded only. For any payment/fee/revenue TOTAL use get_financials — never sum this paginated list. For manual/non-card income use list_income. PLATFORM PAYMENTS — the per-payment fee breakdown shown on the Income page "Platform Payments" table. Each row: date, invoiceNumber, clientName, method (card/etc), status, amount, processingFee (the surcharge the CLIENT paid at checkout), servgroundFee (the platform/ServGround fee the merchant pays), net (what the merchant keeps). All fees are positive magnitudes (costs). Defaults to succeeded payments; pass status to override and from/to to date-filter. For TOTALS use get_financials — do NOT sum this paginated list. For non-Stripe manual income use list_income.

list_payouts

ChatGPT
Stripe Connect payouts + summary (the Payouts page). Returns balance { available, pending } (LIVE account balance, NOT date-filtered), summary { totalPaidOut, inTransit, pending, paymentsReceived }, and payouts[] (amount, status paid|pending|in_transit, arrival/created dates, method). Pass from/to for ANY historical range (Stripe retains full payout history). A payout is a SETTLEMENT (money deposited to the bank) — distinct from revenue: it lags, is net of fees, may be pending/in_transit, so it will NOT equal collectedRevenue. HOW A PAYOUT IS COMPUTED: invoice amount + client processing surcharge − Stripe processing fee − ServGround 1% ≈ payout. For income/revenue/expenses use get_financials instead.

list_permit_cases

ChatGPT
County septic permit cases, newest first: which permit, its PMT number, the property, where it stands (draft, waiting on a seal, ready, filed, approved, denied) and which seals are done.

list_projects

ChatGPT
Project rows (status, value, dates, is_assigned, assigned_technicians[]) or an exact rollup. groupBy=status (count+value per status), groupBy=assignment (assigned vs unassigned count+value), groupBy=technician (workload per field worker; idle techs show 0). Filter by status.

list_proposals

ChatGPT
Proposal rows (status, amount, sent/accepted/rejected dates, client_name) or an exact status/month rollup (groupBy=status|month). Includes client view tracking (first_viewed_at, view_count) — tells whether the client has seen the proposal yet. Use to track the quote pipeline + follow-up.

list_reports

ChatGPT
Report rows (status, published_at, client_name, needs_reply) or an exact status/month rollup (groupBy=status|month). "Sent" reports = status "published"; filter status=published or groupBy=status to count sent reports. Pass needsReply=true to return ONLY active reports with an unanswered client message (excludes archived/withdrawn). Includes client view tracking (first_viewed_at, view_count) and needs_reply=true (an outstanding client message awaits a reply).

list_service_requests

ChatGPT
NOT the services the business offers — for "what services/types of services do you offer" use list_services. groupBy=request_type is the intake KIND (service_request | quote_request | general_inquiry), NOT a service. Service-request rows (status, source, request_type, is_expedited, assigned_to_name, needs_reply) or an exact rollup (groupBy=status|source|request_type|service|referral_source|month|quarter). groupBy=referral_source = how clients heard about the business (unanswered intakes = not_specified). groupBy=month|quarter = intake volume over time. groupBy=service = top / most-requested services by name (NOTE: "service", "service type", and "service_type" all mean the same thing — the service offering). Pass needsReply=true to return ONLY active SRs with an unanswered client message (excludes rejected/cancelled/closed). is_expedited flags rush jobs; needs_reply=true means an outstanding client message awaits a reply. Customer PII is masked when the credential has redaction enabled.

list_services

ChatGPT
USE FOR "what services do you offer", "what types/kinds of services", "service list", "what do you sell" — the services the business offers (pass activeOnly=true). USE FOR the service CATALOG + list prices (what the business offers). For the top / most-requested service by demand use list_service_requests groupBy=service, NOT this. The workspace service catalog (services = service types = the offerings you sell): each service's name, category_name, list price (default_price), is_active, and lead time. Pass activeOnly=true for active services only. Each row also carries audience ("website" = shown to clients; "internal" = hidden, switched off, or a built-in intake type such as General Inquiry) and audience_reason in plain words: answer "what do you offer" from the audience="website" rows first and label any others as internal. NOTE: default_price is a LIST/starting price, not actual billed revenue. For "top / most-requested service (type)", use list_service_requests groupBy=service.

list_team

ChatGPT
USE FOR office/admin staff = workspace users (owner / member / coordinator). For field workers/technicians use list_field_team instead. Office/admin TEAM MEMBERS (workspace users: owner / member / coordinator) — name, role, status, email, and is_technician (true = also a field worker). This is DIFFERENT from field workers/technicians (use list_field_team for those). Pass includePending=true to include unaccepted invites. Staff email is masked when redaction is enabled.

Capabilities

Read

Example Prompts

Click any prompt to copy it.

Similar Apps

Listed on MCP App Store

Listed on MCP App Store

Add this badge to your site or README. It links to this page.

HTML

<a href="https://mcp-app-store.keenethics-labs.com/app/servground"><img src="https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg" alt="Listed on MCP App Store" width="220" height="44"></a>

Markdown

[![Listed on MCP App Store](https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg)](https://mcp-app-store.keenethics-labs.com/app/servground)